2 quite large systetmatic reviews

Introduction

Kambhampati & Vaishya, 2019

Aims

  1. How does quadriceps and hamstrings strength change over time ACLR
    1. compared within-person (i.e. to the uninjured contralateral limb)
    2. compared between-person (i.e. to uninjured healthy populations)

Methods

A group of systematic Reviews with meta analysis

  1. Hip and lower leg strength
  2. Quadriceps and Hamstring strength
  3. Hop performance

Databases: MEDLINE, EMBASE, CINAHL, Scopus, Cochrane CENTRAL, SPORTDiscus

Inclusion criteria: primary ACL injury, aged 18-40 years, with a quantitative measure of quadriceps or hamstring strength

Methodological Quality Assessed on domains outlined by Cochrane Collaboration

Comparison:

  • to the contralateral leg
  • to an uninjured control group

Methods

A group of systematic Reviews with meta analysis

  1. Quadriceps and Hamstring strength

Databases: MEDLINE, EMBASE, CINAHL, Scopus, Cochrane CENTRAL, SPORTDiscus

Inclusion criteria: primary ACL injury, aged 18-40 years, with a quantitative measure of quadriceps or hamstring strength

Methodological Quality Assessed on domains outlined by Cochrane Collaboration

Comparison:

  • to the contralateral leg
  • to an uninjured control group

Effects - Ratio of Means

Effects - Ratio of Means

\[\begin{equation} RoM = \frac{90Nm}{100Nm} \end{equation}\]



Example: ACL Side compared to the contralateral side

ROM = 0.9 95%CI[0.85-0.95]

= ACL side is 0.9x weaker than the contralateral side

= 10% deficit in ACL side strength

The problem

The solution!

Longitudinal/multivariate meta-analysis

  • Ishak et al 2007, Clinical Trials

  • Trikalinos & Olkin 2012, Clinical Trials

  • Cheung 2019, Neuropsychol Rev

  • Mueskiwa et al 2016, PLOS ONE

The solution!

Longitudinal/multivariate Meta-Analysis

Allow multiple (correlated) pieces of information from the same study to be included in a meta-analysis

  • timepoints
    • same people measured over time
  • outcomes
    • same people measured for linked or correlated outcomes (e.g. separate measure for anxiety and depression)
  • comparisons
    • same control groups for different comparators
flowchart TD
  A{Crossley et al} --> B(3 months)
  A--> C(6 months)
  A--> D(12 months)

Traditional “Univariate” Meta-Analysis

  • “2-Level”

  • Random effects for

    • studies (between study heterogeneity)
    • individuals
flowchart TD
  A[Meta-analysis] --> B(Study 1)
  A--> C(Study 2)
  B--> D(i)
  B--> E(iii)
  B--> F(iii)
  C--> G(i)
  C--> H(iii)
  C--> I(iii)

Longitudinal/multivariate meta-analysis

  • “3-Level”

  • Random effects for

    • studies
    • timepoints (or some other clustering variable)
    • individuals
flowchart TD
  A[Meta-analysis] <--> B(Study 1)
  A<--> C(Study 2)
  B--> D(T0)
  B--> E(T6)
  D--> F(i)
  D--> G(ii)
  D--> H(iii)
  E--> I(i)
  E--> J(ii)
  E--> K(iii)
  C--> L(T0)
  C--> M(T6)
  L--> N(i)
  L--> O(ii)
  L--> P(iii)
  M--> Q(i)
  M--> R(ii)
  M--> S(iii)


Data Analysis

  • Mixed-effects meta-analysis with a REML estimator using metafor package.
  • Quadriceps and hamstring, separated by contraction type:
    • Slow concentric
    • Fast concentric
    • Isometric


Random effects:

(timepoint | cohort)

  • timepoints, nested within cohorts

Fixed effect: timepoint

  • linear, log linear, polynomial, 3-knot spline and 4-knot spline

Robust variance estimation methods using clubSandwich package 🥪

Results

  • 233 studies 🥵
  • 🇦🇺 🇬🇧 🇺🇸 🇮🇷 🇯🇵 🇹🇷 🇧🇷 🇳🇴 🇰🇷 🇸🇪 🇨🇭 🇸🇮 🇳🇱 🇬🇷 🇵🇱 🇦🇹 🇮🇹 🇩🇪 🇹🇼 🇳🇴 🇸🇬 🇶🇦 🇨🇳 🇪🇸 🇨🇦 🇩🇰 🇧🇦 🇹🇭
  • 31,234 ACLR participants; 3049 controls; (40% women)
  • Mean age: 18-38 years, median 25
  • Mean BMI: 21.3 to 28.4 kg/m2; median 24.2
  • Most common timepoints:
    • 3 months (k=35)
    • 6 months (k=88)
    • 12 months (k=59)

Quadriceps

Slow Concentric Quadriceps

Fast Concentric Quadriceps

Isometric Quadriceps

Hamstrings

Slow Concentric Hamstrings

Fast Concentric Hamstrings

Isometric Hamstrings

Bonus Analysis!

Exploratory Analysis



  • Several papers report
    • within-person &
    • between-person

Is comparing to the contralateral limb equivalent to comparing
to uninjured controls


Bivariate Analysis

Thank you ++

Functional plots if time

Hop performance

Relationship between different hops